The transformative potential of data and image analysis for eye care

09:00-09:05 Welcome from the Royal Society

09:05-09:15

Abstract

Introduction to the event from lead organiser Professor Emanuele Trucco, University of Dundee

• Listen to the audio (mp3)

09:15-10:00 Gearing an entire country for Health Data Science

Professor Andrew Morris, UK Biomedical Informatics Research Institute, UK

Abstract

Healthcare is arguably the last major industry to be transformed by the information age.  Deployments of information technology have only scratched the surface of possibilities for the potential influence of information and computer science on the quality and cost-effectiveness of healthcare. In this talk, the opportunities provided by computer science and "big data" to transform health care delivery models will be discussed. Examples will be given from nationwide research and development programmes that integrate electronic patient records with biologic and health system data. Two themes will be explored; specifically:

1) How the size of the UK (65M residents), allied to a relatively stable population and unified health care structures facilitate the application of health informatics to support nationwide quality-assured provision of diabetes care.

2) How population-based datasets and disease registries can be integrated with biologic information to facilitate (i) epidemiology; (ii) drug safety studies; (iii) enhanced efficiency of clinical trials through automated follow-up of clinical events and treatment response; and, (iv) the conduct of large-scale genetic, pharmacogenetics, and family-based studies essential for precision medicine.

• Listen to the audio (mp3)

10:00-10:45 How do new imaging technologies and their novel analysis contribute to the understanding of disease? Dr Srinivas Sadda

Abstract

The last two decades, and in particular the last five years, have witnessed an explosion in the development of new imaging technologies. A significant advance is the availability of higher speed, higher resolution, and deeper penetrating swept source OCT technology which allows retinal disease to be studied in an en face plane in a depth-resolved fashion. Adaptive optics techniques have further advanced the axial resolution of our imaging technologies. Other major advances included enhances in contrast, facilitating molecular and functional imaging such as hyperspectral imaging, quantitative autofluroescence, fluorescence lifetime imaging ophthalmoscopy, and OCT angiography.  Mutlidimensional analysis techniques are being developed to cope with these enormous amounts of data. Imaging technology has also benefited from the introduction of widefield devices with specialized mirrors and optics allowing the entire retina to be photographed in a single acquisition. Widefield imaging has provided new insights into various diseases including diabetic retinopathy. Recent studies have highlighted the prognostic importance of peripheral lesions in diabetic retinopathy, with eyes with predominantly peripheral disease having a four-fold higher risk of progression. Combining widefield imaging with artificial intelligence/deep learning approaches, it is now possible to automatically identify and classify diabetic retinopathy, which may have significant public health applications in diabetic retinopathy screening programs.

• Listen to the audio (mp3)

10:45-11:15 Tea break

11:15-12:00 From bench to clinical applications: opportunities in ophthalmic imaging from the lab to the individualised patient therapies - Professor Tunde Peto

Professor Tunde Peto, Queen's University Belfast, UK

Abstract

Ophthalmology has one of the fastest growing imaging arsenal. With these new technologies nearly all components of the human eye can now be imaged. It is a major undertaking to make sure that all imaging modalities are scientifically evaluated for their validity and clinical utility. However, novel clinical imaging modalities raise new, often unexpected questions. These can only be answered by incorporating laboratory research into the evaluations which, in turn, lead to better understanding of the clinical phenotypes. Humans have an incredible ability to recognise patterns and with the new imaging modalities being high enough quality, new patterns are continuously emerging. Therefore, image analysis is still predominantly done manually. As such it is fraught with slow reporting and the need for ever increasing attention to detail.  There is risk of not recognizing crucial features contained in the images leading to the recording of wrong phenotype. Therefore, if we are to cope with the amount of images we generate, there is a need for intelligent, automated, non-human based, though probably human supervised, image analysis.  It is very likely that the development of new algorithms will lead to the development of new treatment paradigms to benefit the patients.

• Listen to the audio (mp3)

12:00-12:45 Retinal ganglion cell degeneration in glaucoma - Professor James Morgan

Abstract

12:45-13:30 Lunch

13:30-14:30 The UK Biobank Eye & Vision Consortium as a model for synergy and collaboration in big data analysis - Professor Paul Foster

Abstract

UK Biobank was initially conceived as a platform for studies of gene environment interaction in major chronic diseases of older age, such as cancer, stroke, MI and diabetes. Towards the end of the study, the UKBB steering committee advised the broadening of the scope of the study to include more detailed examination of participants, including assessment of physical fitness, brain and cardiac imaging, as well as an examination of eyes and vision. Moorfields Eye Hospital and the UCL Institute of Ophthalmology in London developed the eye and vision module. The core funding for the examination was provided by the Wellcome Trust, The Medical Research Council and The Department of Health. Additional support for training, implementation and quality control came from the NIHR Biomedical Research Centre at Moorfields Eye Hospital.
117,649 people took part in the basdeline eye and vision component of UK Biobank, undergoing modified logMAR visual acuity testing on a computerized system developed specifically for UK Biobank, autorefraction and keratometry (Tomey RC-5000), as well as measurement of intraocular pressure on a Reichert ORA Ocular Response Analyser, which returns measures of Goldmann equivalent intraocular pressure (IOP), corneal hysteresis, corneal resistance factor and IOP adjusted for corneal biomechanical properties. A smaller number (68,151) underwent simultaneous colour retinal photography, together with spectral domain optical coherence tomography (SD-OCT) in both eyes. During the data collection phase of UKBB, the Image Reading Centre at Moorfields provided a rapid turn-around quality assurance service for the macular photos and OCT images, finding them to be of high quality, compared with other studies using similar methodology. This large scale, well-curated eye data, together with accompanying extensive phenotyping of all major organ systems, make the UK Biobank eye and vision dataset unique worldwide.
Eye and vision researchers around the UK have formed a consortium involving clinicians and academics from around the UK, incorporating Belfast, Bristol, Cambridge, Cardiff, Dundee, Edinburgh, Gloucester, Leeds, Manchester, Newcastle, Nottingham, Southampton, St Andrews, Warwick and several London centres (Brunel, Imperial, Kings, Moorfields, UCL, St Georges). The consortium interacts using UK Universities JISCMail email list server. The group meets in February each year for a day-long programme of planning, discussion and debate. This has led to the formation of groups working on various aspects of data, including visual acuity, refractive error, intraocular pressure, retinal vascular characteristics, genetics and outcomes adjudication and monitoring. Multiple publications have ensued, with two major genetics studies in press in Nature Genetics. 

• Listen to the audio (mp3)

14:30-15:30 Multi-modal retinal imaging in the Northern Ireland Cohort for the Longitudinal Study of Aging (NICOLA)

Dr Ruth Hogg, Queen’s University Belfast, UK

Abstract

The Northern Ireland Cohort for Longitudinal Aging Study is a comprehensive, long-term epidemiological study of adult development and ageing which started in February 2014 and consists of a random sample of men and women aged 50 years, representative of the Northern Ireland population.  As part of the study approximately 3,600 participants have attended the Wellcome-Wolfson Clinical Research Facility (CRF) at Belfast City Hospital for a health assessment.  The health assessment included anthropometry, respiratory, cardiovascular, cognitive and ophthalmology tests. The ophthalmic component involved visual acuity, auto refraction, corneal compensated intra-ocular pressure, stereo colour fundus photographs, infra-red and autofluorescent retinal images, wide field Optos colour images and Spectral Domain Optical Coherence Tomography (OCT) with Enhanced Depth Imaging (EDI) of the choroid. Macular pigment was also assessed using dual-wavelength autofluorescence. To date most epidemiological studies have only used colour fundus photographs for estimating ocular disease prevalence.  The talk will discuss the opportunities and challenges of using multi-modal imaging and incorporating the data into a larger framework that also encompasses genomic, epigenomic, dietary and biochemical biomarkers.D

• Listen to the audio (mp3)

15:30-16:00 Coffee

16:00-17:00 Facilitated open discussion

Michael Bowen, College of Optometrists, UK

Abstract

Facilitated open discussion: imaging, data and vision- priorities for collaboration, research and eye health.

Led by Michael Bowen, College of Optometrists

• Listen to the audio (mp3)

09:00-10:00 AI in ocular imaging: a China and Singapore perspective

Professor Jimmy Liu Jiang, Cixi Institute of Biomedial Enginerring, Chinese Academy of Science, China

Abstract

In the talk, Jimmy will update the ocular imaging research work conducted in the past years in China and Singapore.

• Listen to the audio (mp3)

10:00-10:45 The QUARTZ retinal image analysis system:quantifying the vasculature in the UK Biobank fundus camera images

Professor Sarah Barman, Kingston University, UK

Abstract

The application of image analysis techniques to provide accurate measures of image features on larger datasets presents specific challenges in ensuring the reliability of the measurements produced. The sizes of current retinal fundus image datasets are increasing, and with their public availability there is increased scope for computer vision researchers to apply algorithms to datasets with variable image characteristics. Image datasets may exhibit variation in terms of image quality, and in terms of image characteristics, such as sets of fundus images that display a range of pathologies.

The presentation will report on the experience of applying computer vision techniques to analyse the UK Biobank retinal fundus dataset that contains over 100,000 images. The analysis included recognition of the retinal vasculature, including classification of arterioles and venules, and the provision of vessel morphometric data across the retinal image. The presentation will discuss the experience in terms of generation of reliable data and will report on the challenges in addition to the overall approach that was taken.

• Listen to the audio (mp3)

10:45-11:05 Tea break

11:05-11:45 Power and limits of automatic retinal image analysis: an image processing perspective

Professor Emanuele Trucco, University of Dundee, UK

Abstract

Automatic image processing holds great promises for the analysis of retinal images from a variety of instruments, and its application to future eye care including the vision of personalized medicine. The introduction of automatic algorithms in diabetic retinopathy screening programs, and the intriguing results achieved by the increasingly ubiquitous artificial intelligence programs are examples of success stories and of potential future achievements. This talk aims to summarise the main methods and challenges of contemporary retinal image analysis in the context of common research and clinical scenarios. Particular attention is given to biomarker discovery, which requires large collections of cross-linked data and lies therefore at the very heart of the theme of this meeting.

• Listen to the audio (mp3)

11:45-12:30 Quantitative retinal biomarkers for diabetic retinopathy: the REVIEW experience

Professor Andrew Hunter, University of Lincoln, UK

Abstract

Diabetic retinopathy results in a complex histopathology in the retinal vessels, with potentially detectable and progressive changes in the vascular geometry (including vessel calibres, tortuosity and branching angles) and resulting lesions (microaneurysms, haemorrhages and exudates). Biomarkers are quantitative indicators that can be reliably extracted from imaging modalities (e.g. fundal images), with minimal or zero user intervention, and are strongly indicative of disease. The challenges in biomarker extraction include: the definition of meaningful quantifiers that summarise usefully often complex and distributed physiological signs, and reliable automated extraction techniques operating in the face of complex, variable physiology and often poor image quality resulting from a challenging problem domain. A key question is: what physiological changes are happening and how can they be elegantly captured? This talk addresses two questions: are there biomarkers allowing early prediction of the onset of diabetic retinopathy; and to what extent does trying to model underlying physiological factors help in biomarker design?

• Listen to the audio (mp3)

12:30-13:15 Lunch

13:15-13:45 Deep learning in the retina

Dr Lilian Tang, University of Surrey, UK

Abstract

Retinal images play an important role in the assessment of many eye and systemic diseases. This talk will give an overview of the developed AI systems using deep learning and image processing techniques for automating the detection of these diseases through digital fundus image data. The talk will also examine the limitation of current approaches on a number of application problems and open up discussion for further intelligence that may orchestrate solutions to various challenges.

13:45-14:30 What can statistics do for the understanding of ophthalmic diseases? From measurement errors and bias to inference and risk estimation using complex data that contain images

Dr Gabriella Czanner, University of Liverpool, UK

Abstract

In the era of collecting complex ophthalmic datasets we are more and more facing the questions of how to best utilise such rich information from patients. In this talk we will discuss three challenges that are sometimes not considered. For example, not considering the measurement errors can lead to lower power of statistical inference. The potential of sector-wise approach to inference from imaging data is often not considered. The accuracy of the traditional statistical approaches to the disease detection is just yet to be fully utilised.

14:30-15:00 Mining large-scale sets of OCT images

Abstract

Ophthalmology is among the most technology-driven of the all the medical specialties, with treatments utilising high-spec medical lasers and advanced microsurgical techniques, and diagnostics involving ultra-high resolution imaging. Ophthalmology is also at the forefront of many trailblasing research areas in healthcare, such as stem cell and gene therapies.         
Moorfields Eye Hospital in London is the oldest eye hospital in the world. Every year, >700,000 patients attend Moorfields - more than double the number of the largest eye hospitals in North America. Together with the adjacent UCL Institute of Ophthalmology, Moorfields is among the largest centres for vision science research in the world. In July 2016, Moorfields announced a formal collaboration and data sharing agreement with DeepMind Health. This collaboration involves the sharing of >1,000,000 anonymised retinal optical coherence tomography (OCT) scans with DeepMind to allow for the automated diagnosis of diseases such as age-related macular degeneration (AMD) and diabetic retinopathy (DR).            

This presentation, will describe the motivation - and urgent need - to apply deep learning to ophthalmology, the processes required to establish a research collaboration between the NHS and a company like DeepMind, the goals of our research, and finally, why I believe that ophthalmology could be first branch of medicine to be fundamentally reinvented through the application of deep learning.

15:00-15:20 Coffee break

15:20-16:30 Facilitated open discussion: Final focus

Michael Bowen, College of Optometrists, UK

Abstract

Facilitated open discussion: Final focus - agreeing activities and actions to address key issues.

Led by Michael Bowen, College of Optometrists 

• Listen to the audio (mp3)

16:30-16:45 Concluding remarks from organisers